AI in hotels: what makes sense today and where the limits are

What AI in hotels can actually do today

When people talk about AI today, they usually mean language models like the ones behind ChatGPT. They understand requests in many languages, recognise what they are about and write a friendly, fitting reply in seconds. What they don’t know on their own: whether your car park costs extra, from when dogs are allowed and whether a double room is still free on the 14th. That knowledge has to come from your hotel system and from information you maintain.

There are also methods that learn from numbers, for example from your booking history. They spot patterns such as “this weekend is booking faster than usual” and can derive forecasts and rate suggestions from them. Here too, quality depends on the amount and cleanliness of your data. A 12-room hotel with one year of booking history delivers less reliable patterns than a 60-room hotel with five years.

No system available today replaces the front desk. The welcome at reception, the solution for the guest with lost luggage and the decision on a goodwill gesture remain human. AI helps where your team currently spends time on the same texts and calculations over and over.

Sensible use cases for small and mid-sized hotels

The overview below ranks typical use cases by benefit and risk. The closer a use case is to money, contracts or personal data, the more important it is for a person to check.

Limits and risks: hallucinations, data protection and labelling

Language models can make up things that sound plausible. Experts call these hallucinations. If a guest asks about an indoor pool your hotel doesn’t have, a poorly set-up AI may still cheerfully give the opening hours. The remedies: the AI may only answer from approved hotel information and real booking data, it should say openly when it is unsure, and a person checks everything beyond simple standard questions. An AI should never change or cancel bookings or promise prices without approval.

Guest messages contain personal data: names, travel dates, sometimes health details such as allergies or information about a disability. If you use AI, you need to clarify which provider processes the data, whether there is a data processing agreement, where the processing takes place and whether data is used for training. Don’t copy guest data into publicly available AI chats in the browser. Use AI functions that are built into your hotel software and covered by contract, and include their use in your privacy notice.

Article 50 of the EU AI Act sets out transparency obligations for AI systems that interact with people; under its timetable they apply from August 2026. In practice this means guests must be able to tell that they are writing to an AI, unless it is obvious anyway. Don’t give an automated assistant a made-up human name with a photo, and label AI replies that are sent automatically. A reply that your team has checked and sent itself is a reply from your hotel. Have the exact obligations for your use case checked by a lawyer.

A person stays in charge of approval

The most important principle for AI in hotels is called “human in the loop”: the AI prepares, a person decides. For guest messages, the AI writes the draft and the front desk reads, adds to it and sends. For rates, the system suggests and management approves. Only once a use case has worked reliably for weeks can individual, narrowly defined cases run automatically, such as the question about check-in time.

Agree as a team who may approve AI drafts, and which topics are never answered automatically: complaints, payments, cancellations, goodwill gestures, health and safety. Check samples regularly, and make sure every automation can be switched off immediately.

How STRYK handles it

STRYK offers two AI add-ons that you activate separately. STRYK Guest AI works in guest communication: it assigns incoming guest emails to the right booking, recognises whether they are about a question, a change or a cancellation, and writes draft replies from your approved hotel information in the language of the enquiry. With a supported, activated OTA connection, it also helps with replies to Booking.com reviews. Guest AI only prepares booking changes and cancellations; a person always has to confirm them. Only standard questions you have explicitly approved are answered automatically, and the automation can be switched off at any time. Guest AI is free for one month and then costs €4 per room per month excl. VAT, billed annually, only if you activate the module.

Intelligent Pricing creates forecasts and rate recommendations per category and stay date from occupancy, pickup, pace, booking curve, lead time, ADR and RevPAR. Every recommendation shows its reasoning; you approve it manually or allow automatic application within your minimum and maximum rates. STRYK does not read competitors’ rates. Intelligent Pricing is also free for one month and then costs 1.9% of the room revenue recorded in STRYK, as long as the module is active. After the trial, neither module continues on a paid basis until you explicitly activate it.

Just as important is what STRYK does not promise: no AI that replaces the front desk, no guaranteed revenue increases and no posting without your review. For supplier invoices, STRYK suggests supplier, amounts and tax rates; nothing is posted until you approve it.